Do you expect the political polls to be scientifically accurate or not

I just visited a popular conservative political blog who's headline was "Even NPR?" in a story about recent polling data. The unspoken message, polls are not scientific especially if the pollster is perceived as being for a particular political camp. Polls are yet another propaganda tool to be used to promote a political agenda. Or WE see polls as dishonest political manipulation fodder for our side and so we by default project that same MO as an assumption of all polls and as long as the final poll before election day is spot on, they can still dishonestly claim impartiality because "after all, they're were right."

Another one is a popular polls averages site. You'd think it would be a place where you could look to to get honest snapshots of where the country is politically with respect to the presidential election safely. After all even if methodology is slightly different (because we all know nobody is deliberately cooking the numbers to make one side look like they're doing better than they truly are) including ALL polls for respected sources and averaging them out would offset any unintentionally inaccurate result. Wait, not so fast...the main poll averages site everybody checks omitted an NBC/Wall Street Journal Poll for over a week that just so happened to look favorable to Obama. Then after other polls came out showing stronger numbers for Romney that offset the NBC/WSJ lead for Obama did they finally include it in their averages.

I would hope the polls were scientifically accurate but I'm starting question them.

Generally they are honest and (reasonably) accurate. There are occasional pollsters who use dishonest methods (e.g. Research 2000) but for the most part they do their best. That's not to say that they are always correct. Built into any poll is their assumption about who constitutes a "likely voter" and which groups are underrepresented via phone interviews. Some pollsters do a better job of this than others...for example, the robopolls tend to be less accurate than those that employ human pollsters, because people are less likely to respond to them and those who choose to respond to them are a self-selecting sample.

With that said, I would expect the final results on Election Night to be within a couple points of the poll averages. It's not unprecedented that the polls are off by more than that (e.g. Harry Reid considerably overperformed his polling average in the 2010 Nevada Senate race), but it tends to be the exception rather than the rule.

As I said before polls like tests are a snap shot in time no matter how good the "science" may be. Whats important is to take the ones that use good methodology and track them over time as a group. This will show real trends. This is essentially what Nate Silver does and so i only look at his "polls".

I am going to think positive here and say yes, I think they will be accurate.

None of the answers listed.

Scientific? Huh? Just like psycobabelism and current "enviro-science" they have some accuracy but are not founded in hard or "real" science. In science, you observe something, create a theory then test it, repeat until theory matches and support observations and you achieve consistent results, then submit to peer review to make sure you didn't accidentally rig the experiments used to test. No way to actually follow this method and come to a scientific conclusion, therefore, they cannot ever be scientifically accurate.

Pole results are too easily skewed. Time of day, location of poll, the wording of the questions, limiting the actual responses to presets instead of what the person actually wants to say, (like the options given for answering this poll) and willingness of people to participate in a poll all affect the outcome. How many people are like me, for example, that don't answer the phone if the number is a toll-free or unknown number (isn't caller id wonderful)? If you poll during working hours, you don't get the workers only those staying at home for some reason (very good method if you want to show liberal results). If you get more urban people participating than rural, or the opposite, then results again skew. Since you actually want results, program your auto-dialer to remember which numbers actually answered and participated then only add in a percentage of truly random numbers for the pole. If you are a biased poller, then you also remember which ones gave the answers you liked and re-call them with only minimal actual randomization to make the numbers look real.

Volunteer polls, like this one, are highly inaccurate because particpation is only those who actually feel like giving a response, if the read the poll at all.

Smeagol said:

I just visited a popular conservative political blog who's headline was "Even NPR?" in a story about recent polling data. The unspoken message, polls are not scientific especially if the pollster is perceived as being for a particular political camp. Polls are yet another propaganda tool to be used to promote a political agenda. Or WE see polls as dishonest political manipulation fodder for our side and so we by default project that same MO as an assumption of all polls and as long as the final poll before election day is spot on, they can still dishonestly claim impartiality because "after all, they're were right."

Another one is a popular polls averages site. You'd think it would be a place where you could look to to get honest snapshots of where the country is politically with respect to the presidential election safely. After all even if methodology is slightly different (because we all know nobody is deliberately cooking the numbers to make one side look like they're doing better than they truly are) including ALL polls for respected sources and averaging them out would offset any unintentionally inaccurate result. Wait, not so fast...the main poll averages site everybody checks omitted an NBC/Wall Street Journal Poll for over a week that just so happened to look favorable to Obama. Then after other polls came out showing stronger numbers for Romney that offset the NBC/WSJ lead for Obama did they finally include it in their averages.

I would hope the polls were scientifically accurate but I'm starting question them.

Click to expand...

If the polls reflect the motivations and work ethic of "the usual call centers," I don't have much faith in them. I think of the people who make these phone calls as the same people who make phone calls to me to buy their products, donate to their causes, etc. I think of the number of times I say, "Thanks, but no thanks," and hang up. Then I wonder how many pollsters are just filling out the responses dishonestly.

I'm sure some polls are accurate; but, as is true with everything on the internet, take them with a grain of salt.

If anyone can find a more accurate poll than Gallup, I hope they'll post up their proof. If I understand their chart, they've been picking them right since 1936:

Election Polls -- Accuracy Record in Presidential Elections

​Thanks for posting this thread. I'm confident in going with Gallup and ignoring the rest.

Edit: For what it's worth, the latest (and, if I understand correctly, the last) Gallup Poll showed Romney up by five points. Gallup has suspended their tracking of the presidential race because of the storm. http://www.enstarz.com/articles/868...shows-romney-leading-5-percent-over-obama.htm

They're BS. Don't trust them implicitly.

MaggieD said:

If the polls reflect the motivations and work ethic of "the usual call centers," I don't have much faith in them. I think of the people who make these phone calls as the same people who make phone calls to me to buy their products, donate to their causes, etc. I think of the number of times I say, "Thanks, but no thanks," and hang up. Then I wonder how many pollsters are just filling out the responses dishonestly.

I'm sure some polls are accurate; but, as is true with everything on the internet, take them with a grain of salt.

If anyone can find a more accurate poll than Gallup, I hope they'll post up their proof. If I understand their chart, they've been picking them right since 1936:

Election Polls -- Accuracy Record in Presidential Elections

​Thanks for posting this thread. I'm confident in going with Gallup and ignoring the rest.

Edit: For what it's worth, the latest (and, if I understand correctly, the last) Gallup Poll showed Romney up by five points. Gallup has suspended their tracking of the presidential race because of the storm. Presidential Polls 2012: Latest Gallup Shows Romney Leading 5 Percent Over Obama : Offbeat : Enstarz

Click to expand...

I don't think Gallup is biased but I am starting to think others are and intentionally so.

The problem with dishonest agenda driven pollsters is they can publish cooked poll results then entire entire election season then a few weeks before election day gradually start bringing them back in line with the accurate data and make the final one accurate then they can claim "We're not biased, we always predict the winning candidate! but they lied their butts off by a few percentage points in order to excite to base, encourage donations, encourage volunteering, etc. falsely showing can win because its neck and neck up until the final poll, the one that is going to be verified by the election results. What I suspect is their dishonesty earlier in the race affecting increased donations, etc. would have been lower and their side would have not been close. I have personally seen one poll averages pollster include daily updates for one poll that consistently favors one candidate, real quick to include polls that favor the same candidate but slow as heck and sometime won't include at all polls that favor the other candidate. Then as long as their final report is honest they can claim bragging rights as being spot on each election and we're all gullible enough to fall for it.

Smeagol said:

I don't think Gallup is biased but I am starting to think others are and intentionally so.

The problem with dishonest agenda driven pollsters is they can publish cooked poll results then entire entire election season then a few weeks before election day gradually start bringing them back in line with the accurate data and make the final one accurate then they can claim "We're not biased, we always predict the winning candidate! but they lied their butts off by a few percentage points in order to excite to base, encourage donations, encourage volunteering, etc. falsely showing can win because its neck and neck up until the final poll, the one that is going to be verified by the election results. What I suspect is their dishonesty earlier in the race affecting increased donations, etc. would have been lower and their side would have not been close. I have personally seen one poll averages pollster include daily updates for one poll that consistently favors one candidate, real quick to include polls that favor the same candidate but slow as heck and sometime won't include at all polls that favor the other candidate. Then as long as their final report is honest they can claim bragging rights as being spot on each election and we're all gullible enough to fall for it.

Click to expand...

Don't forget that polls are a feedback loop for propagandists.

So they can see which persuasive messages are getting traction and which aren't.

ALL campaigns are crafted and managed by PR pros.

Smeagol said:

I just visited a popular conservative political blog who's headline was "Even NPR?" in a story about recent polling data. The unspoken message, polls are not scientific especially if the pollster is perceived as being for a particular political camp.

Click to expand...

When I hear the two (2) terms, "Scientific" and "Political Polls" in the same sentence, I typically chuckle. I used to laugh myself asleep, but that was many years ago. Today, I just smile and continue sipping my Pete's.

There is no such thing as a "scientific political poll" regardless of how the poll might be skewed, Left or Right.

MaggieD said:

If anyone can find a more accurate poll than Gallup, I hope they'll post up their proof.

Click to expand...

In recent years, Gallup's accuracy has been mediocre. Nate Silver has an analysis of Gallup's performance in the recent past, relative to other pollsters. Below are his pollster ratings, and here is his statistical methodology.

(EDIT: Sorry, I'm having trouble uploaded the image showing the rankings. I've replicated the results below. Lower pollster-induced error (PIE) is better.)

Poll Name: PIE
---------------------------
Field Poll: +1.05%
ABC / Washington Post: +1.12%
SurveyUSA: +1.19%
Ciruli: +1.20%
U. Cincinnati / Ohio Poll: +1.25%
Selzer & Co.: +1.31%
NBC / Wall Street Journal: +1.53%
AP-GfK: +1.56%
Mason-Dixon: +1.57%
Pew Research: +1.60%
Gallup: +1.66%
Market Shares: +1.68%
Public Policy Polling: +1.69%
Blum & Weprin: +1.70%
Rasmussen Reports: +1.74%
RT Strategies: +1.75%
Farleigh Dickinson: +1.75%
Rutgers: +1.76%
Marist: +1.77%
Siena: +1.77%
Strategic Vision: +1.83%* (later removed from rankings entirely when SV polls were found to be fraudulent)

If I understand their chart, they've been picking them right since 1936:

Click to expand...

It looks to me like Gallup has had several misses in the nationwide vote: They predicted Bush would beat Gore, they predicted Ford would beat Carter, and they predicted Dewey would beat Truman. Furthermore, several of the elections that they called correctly were blowouts, so they shouldn't get much credit for that. Of the elections where the popular vote margin was 3% or less, they only called 3 out of 6 correctly...no better than a coin flip. But ultimately, I think we should be less concerned with who they predicted would "win," and more concern with their margin of error as a percentage...and this is where Gallup has been mediocre in recent years.

PW4000 said:

When I hear the two (2) terms, "Scientific" and "Political Polls" in the same sentence, I typically chuckle. I used to laugh myself asleep, but that was many years ago. Today, I just smile and continue sipping my Pete's.

There is no such thing as a "scientific political poll" regardless of how the poll might be skewed, Left or Right.

Click to expand...

Then there are the Las Vegas oddsmakers whos livings are determined by a good scientific methodology. A $300 dollar bet on Obama will win you a $100 dollar return. A $300 dollar bet on Romney will win you a $600 dollar return. Why the disparity from the polls in this discussion?

Smeagol said:

I don't think Gallup is biased....

Click to expand...

I'm starting to chuckle again.

There are two (2) bias factors in Political Polling:

a) Source
b) Input

Of course, they are biased. The "Bias" is built into the process of polling itself. Political polls are structurally incapable of producing anything other than biased results.

To explain this, simply take a piece of Enterprise Software. Such software is used by companies and organizations to do many things. A company my want to gain better insight into how it relates to its customers and to provide its customers with a better point-of-contact experience. After going through the RFI/RFP process with several software companies, it makes a decision to invest in a CRM solution.

The Customer Resource Management software is capable of doing a lot of different things to help the company understand its relationship to its customers. One of the "features" of such a solution, gives the company an ability to "Poll Customer Data" contained (typically) inside a DW (Data Warehouse) - a glorified database. The raw polling can be done from within the database, but the CRM solution offers a "ready made" function for doing such polling.

When the company polls that kind of data, it can then use that data for real Scientific purposes: Analysis, Research, Optimization, blah, blah, blah. There is no Source Bias to that kind of data and there is no possibility to produce Input Bias. That's just one example of how you can do real "polling" of a source for use in a Scientific process.

In a political poll, the Source (the polling entity) and the Input (the entity responding to the poll) are subject to bias that comes from cross ideological conflict, because of the way the Source might phrase the query within the poll. So, the problem comes BEFORE the query is ever delivered to the respondent, which sets up one layer of bias, and then the interpretation of the query by the respondent (based on the respondents own ideological bent) sets up a second layer of bias. This process of endemic bias is perpetuated by the fact that precious few political polls ask ONLY one question. Most polls ask a myriad of questions, for which there can be shifting Source bias within each query, leading to overlapping cross ideological conflict in the respondent.

Can you trust these polls? Only to some degree. Their accuracy shifts over time and can become less accurate the closer you get to the actual day of the election. Are these polls "Scientific?" Not, unless you favor altering the definition of what "Science" means. The level of bias that is inherent to the process itself, makes them unlikely candidates for scientific discovery.

However, even a clock is going to be correct twice a day.

I generally trust the Gallup polls.

DaveFagan said:

Then there are the Las Vegas oddsmakers whos livings are determined by a good scientific methodology. A $300 dollar bet on Obama will win you a $100 dollar return. A $300 dollar bet on Romney will win you a $600 dollar return.

Why the disparity from the polls in this discussion?

Click to expand...

- Source Bias
- Input Bias

As outlined above in my post.

No vote, as per usual
All of the choices are true to varying degrees.
I must agree with PW4000.
I think that man is inherently dishonest....the polls are of limited value....

Kandahar said:

In recent years, Gallup's accuracy has been mediocre. Nate Silver has an analysis of Gallup's performance in the recent past, relative to other pollsters. Below are his pollster ratings, and here is his statistical methodology.

(EDIT: Sorry, I'm having trouble uploaded the image showing the rankings. I've replicated the results below. Lower pollster-induced error (PIE) is better.)

Poll Name: PIE
---------------------------
Field Poll: +1.05%
ABC / Washington Post: +1.12%
SurveyUSA: +1.19%
Ciruli: +1.20%
U. Cincinnati / Ohio Poll: +1.25%
Selzer & Co.: +1.31%
NBC / Wall Street Journal: +1.53%
AP-GfK: +1.56%
Mason-Dixon: +1.57%
Pew Research: +1.60%
Gallup: +1.66%
Market Shares: +1.68%
Public Policy Polling: +1.69%
Blum & Weprin: +1.70%
Rasmussen Reports: +1.74%
RT Strategies: +1.75%
Farleigh Dickinson: +1.75%
Rutgers: +1.76%
Marist: +1.77%
Siena: +1.77%
Strategic Vision: +1.83%* (later removed from rankings entirely when SV polls were found to be fraudulent)




It looks to me like Gallup has had several misses in the nationwide vote: They predicted Bush would beat Gore, they predicted Ford would beat Carter, and they predicted Dewey would beat Truman. Furthermore, several of the elections that they called correctly were blowouts, so they shouldn't get much credit for that. Of the elections where the popular vote margin was 3% or less, they only called 3 out of 6 correctly...no better than a coin flip. But ultimately, I think we should be less concerned with who they predicted would "win," and more concern with their margin of error as a percentage...and this is where Gallup has been mediocre in recent years.

Click to expand...

I think we should be more concerned with what the politicians actually stand for and their track record instead of so many simply going with the majority from polls. Overall, I think they really hurt the whole process by convincing some they have already lost, so they don't vote and try to influence those who won't think for themselves to follow the crowd. All political polls distribution should be limited to just the campaigns and never publicly published. That way the campaigns get the benefits and are able to formulate strategy but the voting public is not influenced.